[Intel-gfx] [PATCH v5 1/6] drm/damage_helper: Check if damage clips has valid values

[Mun, Gwan-gyeong]

On Mon, 2020-12-14 at 08:55 +0000, Simon Ser wrote:
> > Userspace can set a damage clip with a negative coordinate,
> > negative
> > width or height or larger than the plane.
> > This invalid values could cause issues in some HW or even worst
> > enable
> > security flaws.
> > 
> > v2:
> > - add debug messages to let userspace know why atomic commit failed
> > due invalid damage clips
> > 
> > Cc: Simon Ser <contact at emersion.fr>
> > Cc: Gwan-gyeong Mun <gwan-gyeong.mun at intel.com>
> > Cc: Sean Paul <seanpaul at chromium.org>
> > Cc: Fabio Estevam <festevam at gmail.com>
> > Cc: Deepak Rawat <drawat at vmware.com>
> > Cc: dri-devel at lists.freedesktop.org
> > Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> 
> After looking at the kernel code, it seems like the kernel already
> checks for
> all of that in drm_atomic_plane_check. Are you aware of this?
> 
> > +	w = drm_rect_width(&plane_state->src) >> 16;
> > +	h = drm_rect_height(&plane_state->src) >> 16;
> 
> The docs say this should be in FB coordinates, not in SRC_*
> coordinates. So we
> shouldn't need to check any SRC_* prop here.
> 
I agree the Simon's opinion. it does check between plane's frame buffer
src geometry and damage clips. (Plane's damage clip might exist outside
of fb src geometry.)